GA4 and Google Ads only give you two real attribution models, not seven. Here's how the data-driven model actually works, the setup step most founders skip, and what this layer still can't see.
THE SHORT VERSION
→Remember the seven attribution models from Part 2? Open GA4 or Google Ads and you’ll only find three, really only two you’d ever choose between: data-driven and last-click. Different layer, not a downgrade, the same ad-platform-versus-CRM split from last time.
→Google’s own guidance recommends data-driven as the default, and it’s been GA4’s default reporting model since the older rule-based models were retired in 2023.
→The single highest-leverage setup step most founders skip: linking Google Ads to GA4. Google’s own data shows linked accounts correlate with a 23% increase in conversions and a 10% reduction in cost per conversion.
→The report to open before cutting any channel: GA4’s Attribution models report (Advertising, then Attribution), which puts last-click and data-driven side by side on the same data and shows exactly which channels last-click was quietly writing off.
→None of this works without consistent UTM tagging on everything that isn’t a Google Ads click. A missing tag doesn’t vanish, it gets folded into "(not set)" or direct, hiding the channel that actually mattered.
→Auto-tagging has to stay on. It’s what lets Google Ads clicks carry a gclid parameter into GA4, and without it, GA4 can’t connect a conversion back to the ad that produced it, no matter which model you pick.
→Attribution isn’t final the moment a conversion fires. GA4 can reattribute a conversion for up to 7 days afterward as more path data comes in, so don’t panic if this week’s numbers shift slightly next week.
→GA4 only sees what happens through Google’s own tracking. Referrals, outbound calls, and dark social never show up here, which is exactly why the CRM-level models from Part 2 still matter for the rest of your pipeline.
A founder opens GA4 looking for the seven attribution models from Part 2, ready to pick the one that fits a five-month sales cycle. Instead there are three options, and two of them are basically the same thing with a different name. That’s not a bug in the platform, and it’s not Part 2 being wrong. Part 2 described how a CRM assigns revenue credit across a whole funnel. GA4 and Google Ads are a narrower tool solving a narrower problem: how paid and organic traffic performs, under a model Google itself trusts enough to make the default. This post is about getting that narrower tool right, because most of the value sitting inside it goes unclaimed simply from a few settings nobody checked.
What GA4 actually gives you to measure attribution
Google’s own Analytics documentation lists the models available in GA4’s attribution reports: data-driven, Google paid channels last click, and paid and organic last click. Data-driven is the default and the one Google recommends. It works by comparing the paths of users who converted against similar users who didn’t, a counterfactual approach, essentially asking how much more likely a touchpoint was to produce a conversion than not having it at all, and assigning credit based on the answer. It’s genuinely closer to the CRM-side algorithmic model flagged in Part 2 than to any of the fixed-rule splits, first-touch, U-shaped, and so on, that HubSpot runs.
One footnote worth knowing before trusting a Monday-morning number: GA4 can reattribute a conversion for up to 7 days after it happens, as more of the user’s path data becomes available. If this week’s channel breakdown looks slightly different next week for the same date range, that’s the model catching up, not the tracking breaking.
The setup step that actually moves the attribution number
Here’s the part most founders skip, and it’s the single highest-leverage thing on this list. Google’s own guidance states that customers who link their Google Ads account to a Google Analytics property are correlated with a 23% increase in conversions and a 10% reduction in cost per conversion. Unlinked, GA4 and Google Ads are each measuring an incomplete version of the same buyer journey, and neither can hand the other the data it’s missing.
Linking takes minutes: in Google Ads, under Tools, Data manager, connect the GA4 property; in GA4, under Admin, Product links, connect the Google Ads account. Once linked, GA4’s key events can become Google Ads conversions, and Google Ads’ cost and click data flows into GA4 reports, so spend and outcome finally sit in the same place.
WORKED EXAMPLE
A $6M ARR B2B company runs Google Ads at a $220 CPA before linking accounts, calculated from Google Ads’ own click and cost data with no visibility into what happened after the click. After linking Google Ads to GA4 and switching the reporting model to data-driven, the same spend shows a $198 CPA, roughly the 10% reduction Google’s own data points to, because GA4 can now see conversions that happened across multiple sessions and touchpoints the unlinked view had been quietly missing or misattributing to the wrong channel.
The ad spend didn’t change. The measurement did. A founder deciding whether to cut or scale that channel is making a materially different decision depending on which number they’re looking at.
Getting the tracking right: tagging
None of the above works if the traffic isn’t tagged consistently in the first place, and there are two separate tagging jobs, not one. Google Ads clicks tag themselves through auto-tagging, which appends a gclid parameter to every ad click, and that parameter is what lets GA4 connect a website session, and eventually a conversion, back to the specific ad, keyword, and campaign that produced the click. Turn auto-tagging off, or let someone turn it off without realizing what it does, and GA4 loses that thread entirely, regardless of which attribution model is selected. Worth a five-minute check in your own account today, because it’s invisible until the day you go looking for a channel’s performance and find it missing.
Everything that isn’t a Google Ads click, LinkedIn ads, email campaigns, newsletters, needs a UTM parameter added manually instead, or GA4 has no reliable way to tell that traffic apart from a direct visit. A campaign missing a UTM doesn’t disappear from the reports, it just gets folded into "(not set)" or direct, quietly inflating a channel that did nothing and hiding one that mattered. Loose tagging produces the same failure mode at every layer of this stack, whether it’s a CRM opportunity with no source field or a GA4 session with no UTM: a model this accurate is only as good as the tags feeding it.
Where to measure it: the Attribution models report
Google’s own documentation points to the Advertising section, then Attribution, where two reports do the real work: the Attribution models report, formerly called Model comparison, and Attribution paths. The first puts two models side by side on the same data, last-click against data-driven is the useful pairing, and shows the percentage swing in conversions and revenue for every channel between them. The second shows the actual sequence of touchpoints a converting user moved through, the closest thing GA4 offers to the conversion path concept from Part 1.
WORKED EXAMPLE
Open the Attribution models report and compare last-click against data-driven for the same 30-day window. A common pattern: paid search shows roughly the same conversion count under both models, because it’s usually the last click anyway, but a display or social channel that looked flat under last-click suddenly shows meaningfully more conversions under data-driven. That gap is the channel’s real assist role becoming visible for the first time, the traffic that introduced the buyer and got quietly written off every month it was judged purely on who closed.
This is the report to open before cutting a channel that "isn’t converting." Last-click will confirm the decision every time, because it’s structurally blind to everything but the final click. Data-driven, sitting right next to it in the same report, is where the actual answer lives.
What this setup still can’t see
Even fully linked, correctly tagged, and running data-driven attribution, GA4 and Google Ads only measure what happens through Google’s own tracking: paid clicks, organic search, and on-site behavior. A referral intro, a founder’s podcast appearance, or a sales rep’s outbound call never generates a gclid or a GA4 session tied to that source unless someone manually tags it. The CRM-side models from Part 2, W-shaped, full path, and the rest, exist precisely to credit the referral partner and the podcast clip that this layer was never built to see.
Frequently asked questions
Does the Model Comparison report replace what my CRM shows me?
No. It only compares models within GA4’s own paid-and-organic data, so it will never surface a referral, a podcast listener, or an outbound reply the way a CRM’s W-shaped or full path model would. Use it to sanity-check a channel-cut decision on the paid side; keep the CRM as the source of truth for the deals themselves.
I linked Google Ads and GA4 months ago. Do I need to check anything else?
Yes, check that auto-tagging is still enabled and that the GA4 property is still receiving the linked account’s data under Admin, Product links. Links can quietly break after a Google Ads account restructure, an agency handoff, or a permissions change, and the first sign is usually a channel’s numbers looking wrong for reasons nobody can immediately explain.
Why does my Google Ads dashboard show different conversion numbers than GA4?
Almost always a model mismatch: Google Ads may report on its own conversion tracking while GA4 reports on a different attribution model or a different conversion window. Confirm both platforms are looking at the same underlying key events and the same attribution model before assuming either number is wrong.
Should I switch from last-click to data-driven right now if I haven’t already?
The one real exception is a brand-new account without enough conversion volume yet, where data-driven has too little path data to meaningfully differ from last-click, worth revisiting once volume grows. Outside that case, Google’s own recommendation and default setting both point the same direction.
How do I actually choose between the seven CRM models covered in Part 2?
That’s the decision framework still owed: how sales cycle length, deal size, and buying committee size point toward one CRM model over another, plus how to run two models side by side before committing to one for board reporting. That’s next.
Linking, tagging, and reading the Attribution models report correctly is table stakes. It answers one question well: how is paid and organic traffic actually performing. The referrals, the podcast clip, and the seven-model decision from Part 2 still live one layer up.
THE DIAGNOSTIC
Is your GA4 setup actually feeding you the truth?
One question. One personalised recommendation for your tracking setup. No sales call required.